NERD

What is NERD?

NERD (Net Estimated Rating, Denoised) is our all-in-one player impact metric: an estimate of how many points per 100 possessions a player adds to his team, split into offense (O-NERD) and defense (D-NERD). It combines two kinds of evidence, each weighted by its reliability: what actually happens on the scoreboard when the player is on the floor (a multi-year, luck-adjusted RAPM in which every player in the league is estimated jointly), and what his statistical profile suggests (a box-score prior fit against long-run impact). Thin samples lean on the profile; large samples earn their own number.

The Denoised part is what makes it ours: every possession behind NERD runs through this site's luck-adjustment pipeline — opponent three-point variance and other scoreboard noise are muted before the model ever sees the data, and all stint scores are calibrated to official box scores. Playoff possessions are included in the on-court evidence; the statistical profile uses regular-season stats only.

PAR and WAR convert the rate into season totals against a replacement baseline (−2.0 per 100 — a full team of freely-available players, which our win model puts at about a 16-win season): PAR = (NERD + 2) × possessions / 100 is total points added over a replacement player across the season, and WAR = PAR / 28.8 converts points to wins. NERD itself stays centered on league average (0 = average player); PAR/WAR reward both quality and durability. In Career mode they sum across seasons.